Enhancing Generalization in Sickle Cell Disease Diagnosis through Ensemble Methods and Feature Importance Analysis

TL;DR

The paper addresses the challenge of generalization in SCD diagnosis from peripheral blood smear images by designing an ensemble-based classification pipeline and conducting a thorough feature-importance analysis for interpretability. It leverages a diverse set of classifiers (DT, ET, RF, GB, SVM, kNN, MLP) and two fusion strategies ( voting and stacking ) on a feature set comprising $41$ shape, $62$ texture, and $18$ color features, totaling $121$ features, with all features standardized before modeling. The best generalization performance on a new dataset is achieved by a stacked ensemble of RF and ET, delivering a $F1$-score of $90.71\%$ and an $SDS$-score of $93.33\%$, outperforming previous state-of-the-art generalization results ($F1$ around $86-87\%$, $SDS$ around $89-89.5\%$). The study also identifies the most informative features to reduce complexity and training time, and provides open-source code, model parameters, and data to support reproducibility and real-world diagnostic deployment.

Abstract

This work presents a novel approach for selecting the optimal ensemble-based classification method and features with a primarly focus on achieving generalization, based on the state-of-the-art, to provide diagnostic support for Sickle Cell Disease using peripheral blood smear images of red blood cells. We pre-processed and segmented the microscopic images to ensure the extraction of high-quality features. To ensure the reliability of our proposed system, we conducted an in-depth analysis of interpretability. Leveraging techniques established in the literature, we extracted features from blood cells and employed ensemble machine learning methods to classify their morphology. Furthermore, we have devised a methodology to identify the most critical features for classification, aimed at reducing complexity and training time and enhancing interpretability in opaque models. Lastly, we validated our results using a new dataset, where our model overperformed state-of-the-art models in terms of generalization. The results of classifier ensembled of Random Forest and Extra Trees classifier achieved an harmonic mean of precision and recall (F1-score) of 90.71\% and a Sickle Cell Disease diagnosis support score (SDS-score) of 93.33\%. These results demonstrate notable enhancement from previous ones with Gradient Boosting classifier (F1-score 87.32\% and SDS-score 89.51\%). To foster scientific progress, we have made available the parameters for each model, the implemented code library, and the confusion matrices with the raw data.

Figures (4)

• Figure 1: Examples of the three types of red blood cells to take into account in SCD diagnosis.
• Figure 2: Research methodology outline.
• Figure 3: Circular cells classified as elongated or other.
• Figure 4: Elongated or other cells classified as circular.